Home // SPWID 2023, The Ninth International Conference on Smart Portable, Wearable, Implantable and Disability-oriented Devices and Systems // View article

Skin Lesion Segmentation and Classification Using Deep Learning Methods

Authors:
Spyridon Vollas
Isidoros Perikos
Michael Paraskevas

Keywords: melanoma; deep learning; ensemble learning; convolutional neural networks; capsule networks.

Abstract:
Melanoma is the most dangerous type of skin cancer. Every year hundreds of thousands of people are affected by this form of cancer, with tens of thousands of them leading to death. The diagnosis at an early stage is crucial and can lead to a complete cure of patients making it the most important parameter in the fight against it. The aim of this paper is to design and formulate deep learning methods and an ensemble schema for the accurate recognition of melanoma as well as of other skin lesions. The methods and the deep learning architectures that were designed and tested are convolutional neural networks, deep convolutional neural networks, deep residual networks as well as capsule networks. An ensemble method which consists of the DesNet121 and ResNet50 architectures is also designed and introduced. For the DenseNet121 and ResNet50 methods, the transfer learning technique was used for the phase of training. The experimental study revealed quite interesting results on the HAM10000 and ISIC 2019 datasets. The best performance among the methods was achieved by the DenseNet121 network with an accuracy up to 81%.

Pages: 5 to 10

Copyright: Copyright (c) IARIA, 2023

Publication date: June 26, 2023

Published in: conference

ISSN: 2519-8440

ISBN: 978-1-68558-075-9

Location: Nice, France

Dates: from June 26, 2023 to June 30, 2023